The problem of judgement in constructing numerical models for policy analysis

In many different fields it is now commonplace for complex numerical models to be constructed with a view to illuminating public policy issues. The an alysts who construct these models have to make many decisions on how to rep resent the issues numerically, and they sometimes seek expert advice in thi s task. In this paper we first justify our assertion that modelling problem s of this nature involve judgement. Then we discuss the problem of how to d etermine appropriate expert judgement for this task. Finally we argue that the status of expertise in this task is problematic and that this has impli cations for interpreting the outputs of such models.